Extended Data Fig. 6: Tumour niche fibroblasts transition towards Cancer Associated Fibroblast (CAF) identity; (Corresponds to Fig. 3).

a, Heatmap showing the log₂ fold-change (FC) of differentially expressed genes in Pdgfra^low relative to Pdgfra^high fibroblasts. Genes characterising Pdgfra^low shown in red, Pdgfra^high in blue. DAPI (blue). Scale bars, 10 µm. Dashed line marks fibroblast contour. b, UMAP projection denoting the heterogenous expression of Fn1 (like Pdgfra) in fibroblasts; inset shows fibroblast cluster distribution. c, Violin plots showing the levels of Pdgfra and Fn1 expression in fibroblast clusters i.e., 3, 2, 4, and 19 in control conditions. Fn1 expression levels are inversely proportional to those seen of Pdgfra. Black line across violin plots shows is mean. d, Representative images of stromal whole mount showing fibronectin (FN1, green) expression in lamina propria PDGFRα^low (white arrowhead) and submucosae PDGFRα^high fibroblasts, labelled in yellow (from induced Pdgfra^CreERRs26^FConfetti mice; EDF 4j; n = 3 mice). e, Stacked bar plot (left) showing fibroblast cluster enrichment across conditions. Significance was assessed by one-sided Chi-squared test, standardised residual (Std Res) values are shown in the heatmap on the right, indicating Cluster 19 (C19) to be the most enriched cluster in tumour conditions f, Volcano plot showing differential gene expression between Tumour and DEN conditions in Cluster 19, Pdgfra^low fibroblast cluster enriched in tumours. Non-significant genes (p ≥ 0.05), green and grey. Significant genes with FC > 1, red; FC ≤ 1, blue. Genes defining the profibrotic nature of cluster 19 in tumours are labelled on the right. g Representative confocal images of a DEN adjacent area and nascent tumours 10 d after DEN treatment or persisting tumour 6 m after DEN treatment showing the accumulation of Fibronectin (FN1, green). DAPI (blue). Scale bars, 25 µm (10 d), 100 µm (6 m). h, Second harmonic generation (SHG) image of lamina propria (directly underneath the epithelium) in a control (Ctrl) and tumour areas 3 m post-DEN. Dashed lines mark insets. Extracellular matrix (ECM) fibres detected by SHG, (cyan); nuclei (red). Scale bars, 25 µm. The yellow diamonds represent the longitudinal orientation of the oesophagus. i, ECM fibre density was scored as SHG signal intensity 3 m after DEN treatment. N is the number of regions or tumours assessed from 6 animals; n = 36 (Ctrl), 20 (DEN), 31 (Tmr). Bars indicate mean±s.e.m. One-way Welch’s ANOVA with multiple comparison was used to assess significance. j, Plot depicting the orientation of ECM fibres in control regions and tumours 9 m after DEN treatment. N is the number of fields of view assessed in 3 animals per condition, n = 12 (Ctrl) and 8 (Tmr). Data expressed as average ±s.d. k, Representative confocal image of a nascent tumour from 6 animals, showing integrin α6 (CD49f, magenta) sequestering in the early tumour niche. Samples were also labelled for KRT6A (red), PDGFRα (grey) and DAPI (blue). Image settings adjusted to upper stromal layer. l, Venn diagram displaying overlapping genes between known CAF markers (top left, red) and differentially expressed genes (DEGs) in DEN tumour (Tmr) fibroblasts (Cluster 19) relative to external control (Ctrl; top right, yellow) or to internal control cells (DEN; bottom, green). Overlapping genes are listed. Known subtypes of CAFs: Myofibrotic (my), immune (i), and antigen-presenting (ap) were considered. m,n, Violin plots showing expression of signature genes found to be enriched in Cluster 19 tumour fibroblasts (m) or other canonical CAF markers (n); external control (ctrl, green), internal control (DEN, blue), and tumours (Tmr) (red). Black line is mean. o, Representative 3D-rendered side-view images of control tissue and tumours 6-7 m post-DEN. Expression of CAF markers (FAP, VIM, FSP) was not detected in niche fibroblasts. KRT6A (red); CD45 (immune), and CD31 (endothelia) (cyan); Vimentin (VIM) (green); PDGFRα (top left and right) and FSP (bottom), (grey); FAP (green), DAPI, blue. Scale bars, 50 µm. Dashed line outlines tumour area. p, Quantification of Ki67+ EGFP+ fibroblasts from Pdgfra^EGFP mice in control and Niche+ tumours 6-12 m after DEN treatment. N = 25 control areas (from 7 animals) and 50 tumours (from 8 animals). Data is expressed as mean±s.e.m. Two-tailed Mann Whitney test. q, Representative bright field (H&E) and confocal (IF) images of invasive carcinoma from 14 m, post-DEN withdrawal tissue. H&E was used to diagnose the pathology. In different panels αSMA, FAP, VIM (green) co-expression with PDGFRα (grey) show CAF emergence in DEN model. DAPI (blue). Scale bars, 100 µm (black) 50 µm (white); representative data from 5 animals >12 m after DEN treatment. r, Violin plots showing the prediction score from label transfer for Pi16+ and Col15a1+ cross-tissue universal fibroblast population⁴⁰ in upper GI fibroblast clusters. s, upper GI fibroblast clusters (top left). Remaining UMAPS show Pdgfra expression in UMAP space of upper GI and cross-tissue fibroblasts⁴⁰. Days (d), months (m). Source data (i,p).

Source data